Voiceprint recognition technologies are also called speaker recognition technologies, and are technologies that distinguish speakers according to voices and are used for identifying identities of the speakers. Generally speaking, speaker recognition technologies mainly include target speaker model registration and score measurement, so as to obtain a recognition result according to an obtained score. For example, in a Gaussian Mixture Model (GMM)-Universal Background Model (UBM)-based system, first, a UBM and a GMM of a target speaker are used to obtain a score function of a characteristic parameter of a test voice based on a log-likelihood ratio; then, the characteristic parameter of the test voice is taken as an input of the score function to obtain a corresponding speaker score; if the score is higher than a preset threshold value, a recognition result is true; if the score is lower than the preset threshold value, the recognition result is false.
However, the method for speaker recognition in the prior art has low identifiability.